On Fri, Jun 10, 2005 at 07:22:14AM +0000, Johnny C. Lam wrote:
> On Fri, Jun 10, 2005 at 07:54:16AM +0100, Alistair Crooks wrote:
> > On Fri, Jun 10, 2005 at 03:49:08AM +0000, Johnny C. Lam wrote:
> > > On Fri, Jun 10, 2005 at 01:37:21AM +0000, Dan McMahill wrote:
> > > > 
> > > > Log Message:
> > > > add autoconf test for missing (like in openbsd-3.6) or broken (like
> > > > in AIX-4.3.2) PRIu64.  If it is missing or broken, provide one.  Should
> > > > address PR30375.  Welcome to pkg_install-20050607
> > > 
> > > This should *not* have caused a version bump.  There is *no* functional
> > > change here.  Please back out the version bump.
> > 
> > What's the problem with the version bump? I think it should be there -
> > that way, we can identify whether it will work on AIX and OpenBSD. Also,
> > the way the dependencies work within pkgsrc means that the pseudo-date
> > version should be bumped whenever anything changes in pkg_install (it
> > used to be calculated directly from CVS Ids, after all).
> 
> This fix falls under the "fix broken compile on platform X" and has
> historically never led to a version bump.  On systems that already
> have pkg_install-20050530 installed, there is no difference between
> it and a pkg_install with Dan's fix.  On OpenBSD and AIX, since it
> wasn't possible to install pkg_install-20050530 prior to Dan's fix,
> the only relevant fact is that it's possible to install it now with
> an updated pkgsrc tree.

Indeed.

All of which is a perfectly valid reason to bump the version number.

It used to be bumped automatically whenever any source file under
pkg_install was checked in, and I really can't see what the problem is
to do exactly the same thing here.

What is the limitation or restriction on bumping the version number?

Regards,
Alistair